Re: On metrics

[Phil Davis <pmd8@cornell.edu>]

Dear Aaron,

Thanks for your thoughtful response to the list.  COUNTER 
compliance could in the future require publishers to adhere to 
filtering out industry-defined lists of robots.  But as you 
rightly described, there are many unregistered robots, and 
figuring out whether a machine or a human being is behind an IP 
address is becoming more difficult to do.

I'm still a bit confused with the intention of bepres's news 
release. Is the purpose simply to notify subscribers on your 
changes and their impact on download statistics? Or is there an 
intent to push the industry to adapt to a more rigorous (and thus 
more valid) set of standards?  If the latter is the case, this 
puts bepress in a bit of a dilemma.  By publicizing your new 
filter, you release the blueprint of how one could code the 
behavior of a robot to actively avoid being filtered.  At this 
point, details are left conspicuously vague [1] and subscribers 
will just need to take your word that you are attempting to 
report honest figures.

Secondly, it seems that there is an underlying value frame to 
your arguments.  For instance, someone who downloads only one 
article for the purposes of reading is counted whereas another 
who downloads a hundred in order to do computational linguistics 
is not.  Humans are valued through their intentionality, but 
robots are not.  Yet robots do provide a real value to potential 
readers since they actively index and provide entry points for 
access; or by downloading and caching, robots can provide 
redundancy and archiving.  In sum, deciding what goes into one's 
filter makes implicit value judgments on what is worth (or not 
worth) counting.


[1] "Remove downloads from unidentifiable RACS, using Bayesian 
analysis and other heuristics" from: 
http://www.bepress.com/download_counts.html


-- 
Philip M. Davis
PhD Student
Department of Communication
336 Kennedy Hall
Cornell University, Ithaca, NY 14853
email: pmd8@cornell.edu
https://confluence.cornell.edu/display/~pmd8/resume